The paper is concerned with the requirements of expert systems for providing advice to persons (so-called users) who are faced with acute problems of practical, everyday life. The main contribution of the study is a review of observations made of actual conversations between users and expert advisors.

The main conclusions are that users seeking practical advice are primarily concerned with remedies, not with identifications of faults, and that the proper remedy can only be found by the expert who is guided by the user’s specific intentions, expectations, and constraints, as brought up in the conversations. What is not noted is that these conclusions are obvious as soon as it is realized that the very notion of a problem implies a person in a problem situation, and that any problem must include, as an inherent aspect, that person’s situation, attitudes, experience, etc.; in other words, items that are infinitely varied and can be brought out only by the person himself or herself. It is realized in the paper that artificial response systems that take these aspects into account in full generality “will not be achieved for some time yet.” Meanwhile, the authors are engaged in experiments on less ambitious formulations of the advice-giving problem.